dtcxzyw / llvm-opt-benchmark

An LLVM IR dataset for data-driven compiler optimization research
https://dtcxzyw.github.io/llvm-opt-benchmark/
MIT License
22 stars 6 forks source link

Update diff September 30th 2024, 1:37:05 pm #1399

Open github-actions[bot] opened 1 month ago

github-actions[bot] commented 1 month ago

from: https://github.com/llvm/llvm-project/commit/1efd1227b2042b865afb7c5a260f2d96927bf911 to: https://github.com/llvm/llvm-project/commit/5df7d88c37877756984a533e4b5ccaad015c7491 commit: c2f0f5ad93c08d4eb27f3f191d4ea6f4f2f59127

Change Logs

from 1efd1227b2042b865afb7c5a260f2d96927bf911 to 5df7d88c37877756984a533e4b5ccaad015c7491

5df7d88c37877756984a533e4b5ccaad015c7491 [gn build] Port 8e6bba230e1c 7061d386f19aa389d405ad791b2ca4adf3946d9d [gn build] Port 18df9d23ea39 3d9cf8b397f61f6259049a625f4888a119352e0d [clangd] Improve filtering logic for undesired proto symbols (#110091) 18df9d23ea390eaa50b41f3083a42f700a2b0e39 [libc++] Add an ABI setting to harden unique_ptr<T[]>::operator[] (#91798) 8e6bba230e1c079f1fcc3cc6f565e1a557f34379 [libc++][NFC] Rename fold.h to ranges_fold.h (#109696) 735a5f67e351fc072a9f7c18b030036681f7935a [AMDGPU] When allocating VGPRs, VGPR spills are not part of the prologue (#109439) c0e97c4dfc936ed13ae9696f0615d7e648268bc9 [Clang][LLVM][AArch64] Add intrinsic for LUTI4 SME2 instruction (#97755) (#109953) 587eaefe876f8fd2cd8b9b0862d81f35b7f9ebce [sanitizer_common] Handle ptrace on Linux/sparc64 (#109310) 412d59f0a510a05c08ed45545943dfd2f901bc5d [DAG] combineShiftToMULH - handle zext nneg as sext bdd35593059441dd626305ee48a91d19b952189c [X86] pmulh.ll - add extra test coverage from #109790 82e594a022d3f3e19ba22dfc55349fe97ed844e6 [AMDGPU] Remove unused lambda capture in AMDGPUMCResourceInfo.cpp (NFC) 8f50dbd7b8a2ed7206d0ad4c90129c014eb032fb [gn build] Port c897c13dde3b c897c13dde3bb413e723317c0579781fb6045a8b [AMDGPU] Convert AMDGPUResourceUsageAnalysis pass from Module to MF pass (#102913) 3c85102756ca0737e2282794451889cac502389f [clang][bytecode] Handle DiscardResult for fixed-point literals (#110475) 79382eb97adff71dbc9d4969431dd4a0967ee7ad [clang][bytecode] Remove superfluous check from fixed-point div (#110478) dc6e4805a0a563052c0c8d81628718fcd6acb193 [clang][bytecode] Cast fixed-point cmp result to int if necessary (#110469) 9f3728d15721830ef1d29a8ae1af97124bc26b47 [libclc] Fix installation w/ ENABLE_RUNTIME_SUBNORMAL (#109926) 81ba95cefe1b5a12f0a7d8e6a383bcce9e95b785 FastISel: Fix incorrectly using getPointerTy (#110465) 5883ad34d66075fdb45ee2544a8bede69a13053b DAG: Handle vector legalization of minimumnum/maximumnum (#109779) d556e38fe8062d4c03258bd268f62afee10aaebe [flang][debug] Support derived type components with box types. (#109424) 6f956e3117589e6a4c71ed7f27095aa232778304 [AMDGPU] Rename LocalMemorySize features to AddressableLocalMemorySize (#110242) b8c974f09391d78035928c599a911009bbe49e85 [MLIR][TilingInterface] Extend consumer fusion for multi-use of producer shared by terminator ops (#110105) 8e0daabe97cf5e73402bcb4c3e54b3583199ba8f AMDGPU: Make a frame index test more realistic 83fe85115da9dc25fa270d2ea8140113c8d49670 AMDGPU: Fix inst-selection of large scratch offsets with sgpr base (#110256) 93eaa992896d74eddaf51e8e2601fabc41a1e316 [abi] [ItaniumMangle] Remove a test case that fails due to expected r… (#110467) e9d12a6b451bd403d95105aa976a011dc821f126 AMDGPU: Add test for 16 bit unsigned scratch offsets (#110255) f627c453dbef21d6949dabbce2e7a424b9deda30 Fix test for PR#109680 f445e39ab271d07733f0f45048badd9e58905aec [SimplifyCFG] Use isWritableObject() API (#110127) dd2792ac7d611fbf952bc0d240fc98170f9f8b4f [bazel] Fix build past 6292f117c39b9fc72da4e40328eeeda2aa94a5f2 (#110459) a548eded70dd92296439bd8561dbeeb67fc76711 [AArch64][SME]Check streaming mode when using SME2 instruction in fra… (#109680) f5c02dd06e03969e8d95746fc184609fe28f509d [MemCpyOpt] Use EarliestEscapeInfo (#110280) f8373cb0f91c63ec3f1bd887ccae4387d9cd0365 [LV] Reuse VPReplicateRecipe to handle scalar stores in exit block. (#106342) 147558e31c2b8d2986fc13cf6f9e2bf6a120f2e6 [clang][ItaniumMangle] Mangle friend function templates with a constr… (#110247) 0d384fe978f9675be9fe960940daa2fd599104ed [clang][analyzer] Move 'alpha.core.PointerSub' checker into 'security.PointerSub' (#107596) a87640c97e5ce5b88304b4149c0e59f4399aff0a AMDGPU: Fix assertion on load of vector of pointers (#110436) af47038fb1385435eb315cc1962464f19ea9e186 [clangd] [C++20] [Modules] Support code complete for C++20 modules (#110083) 0bc98349c84130babb6a4acf2951a0ccc84f574f [LICM] Use DomTreeUpdater version of SplitBlockPredecessors, nfc (#107190) 6f3c15163f9e8852768f0dd121564bffe0ace6a6 [SPIR-V] Fix of OpString separator in DI test (#110249) e20bf28987b74ed4f4b48f49f4506d0659c09bed [HLSL] Replace element_type* handles in HLSLExternalSemaSource with __hlsl_resource_t builtin type (#110079) 6292f117c39b9fc72da4e40328eeeda2aa94a5f2 [ORC-RT] Rename sections_tracker.h to record_section_tracker.h. 4ef77d61b2ee3054344b50d5f4e3111ce69fffcf [include-cleaner] Attach Header to AnalysisResults for misisng headers (#110272) 64f2bff12b8ac40c79004ffacf46a5294600d219 [ReachingDefAnalysis] Turn MBBReachingDefsInfo into a proper class (NFC) (#110432)

Compilation time result (by files):
Top 5 improvements:
  xgboost/c_api.cc.ll 42736262794 -> 40572106472 -5.06%
  meilisearch-rs/4rtt9oltj0ubuf08.ll 28866177269 -> 27815596620 -3.64%
  image-rs/2s4mh02dvph60euq.ll 11134966689 -> 10799338848 -3.01%
  delta-rs/1q6q2cmruh1j9ndm.ll 727728987 -> 712502364 -2.09%
  ockam-rs/3aq45j75ob7wj4as.ll 1364563692 -> 1340065158 -1.80%
Top 5 regressions:
  php/parse_date.ll 16791032166 -> 46194683780 +175.12%
  wasmtime-rs/beiugzjs7tvizh4.ll 211748684 -> 292958802 +38.35%
  wasmtime-rs/2mzhr81bqzbenmxl.ll 277930750 -> 300395912 +8.08%
  wasmtime-rs/1gzboj2l8wur0jva.ll 143799287 -> 155250360 +7.96%
  wasmtime-rs/3wjnkzkdrv5snplq.ll 4419050738 -> 4740794539 +7.28%

Overall: -0.01729196%
Compilation time result (by projects):
Top 5 improvements:
  xgboost 409367231116 -> 407058869422 -0.56%
  image-rs 169662107019 -> 169358202704 -0.18%
  g2o 470528922035 -> 469711793505 -0.17%
  tinympc 19054415913 -> 19027997748 -0.14%
  taskflow 275105541409 -> 274799484484 -0.11%
Top 5 regressions:
  php 409340733935 -> 438971721901 +7.24%
  pyo3-rs 26479920696 -> 27040867190 +2.12%
  logos-rs 10263690879 -> 10336551948 +0.71%
  yyjson 16946291319 -> 17055045097 +0.64%
  pugixml 9890300801 -> 9950099363 +0.60%

Overall: 0.03900368%
Improvements:
capture-tracking.NumNotCapturedBefore 8015177 -> 9480749 +18.28%
capture-tracking.NumCapturedBefore 8428943 -> 9234596 +9.56%
memcpyopt.NumMemCpyInstr 622645 -> 642166 +3.14%
dse.NumFastOther 104741 -> 105751 +0.96%
memcpyopt.NumStackMove 38103 -> 38309 +0.54%
memcpyopt.NumCpyToSet 8499 -> 8542 +0.51%
dse.NumFastStores 588867 -> 591769 +0.49%
bdce.NumSimplified 6017 -> 6032 +0.25%
memory-builtins.ObjectVisitorLoad 1569940 -> 1573342 +0.22%
dse.NumGetDomMemoryDefPassed 749660 -> 751275 +0.22%
Regressions:
memcpyopt.NumMoveToCpy 6361 -> 6293 -1.07%
dse.NumCompletePartials 18954 -> 18900 -0.28%
simple-loop-unswitch.NumTrivial 2129 -> 2123 -0.28%
correlated-value-propagation.NumNonNull 5458541 -> 5443725 -0.27%
sroa.NumAllocaPartitionUses 169145821 -> 169058456 -0.05%
licm.NumMovedCalls 15250 -> 15243 -0.05%
bdce.NumRemoved 266603 -> 266497 -0.04%
sroa.NumAllocaPartitions 57306252 -> 57284454 -0.04%
gvn.NumGVNLoad 869311 -> 869014 -0.03%
sroa.NumAllocasAnalyzed 76156190 -> 76131486 -0.03%
dtcxzyw commented 1 month ago

See https://github.com/dtcxzyw/llvm-opt-benchmark/pull/1387#discussion_r1779353666